Communicating with Data

Chapter 1. Communication

Trying to get anything done in an organization always requires more communication than you first expect. On any given day, you might be communicating with the following:

• The executive, to ask for funding to support a new project

• Your management, getting agreement to use their people’s time

• Your peers, to deal with new challenges that arise daily

• Your customers and clients, to deal with their latest requests

• Your suppliers, to ensure that your logistics chain is ready

• Your team, to change priorities as needed

If everyone else in your organization is communicating this much, that’s a lot of activity. Trying to make your points and questions heard—or even just reading, organizing, and responding to this onslaught—is a tough challenge for anyone. Any advantage you can gain will make you more effective at your job. This chapter shares the key aspects of communication by applying them to data visualization and shows you how understanding these basic principles can help you communicate more effectively.

You need to be heard, but you also need to ensure that what you say makes an impression. To help with that, this chapter also discusses the final, commonly overlooked part of the communication process: the receiver must retain the information you communicate in their memory.

The aim of this book isn’t to teach you something new about your area of expertise but rather to help you share your knowledge more effectively. One way to do that is to combine it with data to validate your opinion. The engineer W. Edwards Deming is often quoted as saying, “Without data, you’re just another person with an opinion.”

Deming was one of the developers of Total Quality Management, a management framework that focuses on improving processes to create better and more consistent outputs. If you’re suggesting ways for your organization to improve what you do and how you do it, you should back up your arguments with data.

If you are new to working with data, all this might feel imposing. The good news is that working with data is not as intimidating as it seems. Chapter 2 gets into the details of working with data, but first I want to introduce you to what makes data visualization so effective.

What Is Communication?

Good ideas are useless unless you can get other people to understand them. Getting people to understand your point of view takes careful communication. But what do I mean by communication?

The Communication Process

Communication is something you do without thinking about it every day. You share thoughts and ideas with others by speaking, writing, or just using expressive body language. What you are subconsciously doing with lots of your communication is creating a message and sending it to the person you hope receives it.

The act of sending and receiving a message is only part of the process: you encode the message in a way that you think will be clear to the receiver—that is, they will be able to decode it, or understand what you are trying to tell them. The sociologist Stuart Hall describes this process in his classic work “Encoding and Decoding in the Television Discourse”. Hall describes how these concepts work in television media; you can apply a similar approach to your own communications. However you want to communicate with others, you are choosing how to take the information you have and share it. The method that you use to share it will require you to encode your thoughts. Therefore, your audience will need to decode the message to understand exactly what you meant.

Another factor in communication I often think about comes from a mathematician writing about passing messages through limited bandwidths. In 1948, Claude Shannon described communication in a way that’s still relevant today, and ever since I saw it, I think about it in regards to data visualization. I’ve updated Shannon’s original diagram here to focus specifically on personal data communication in the way that I think about it (Figure 1-1).

Figure 1-1. The data communication process in organizations

Let’s look at how Shannon’s model translates to everyday communication within an organization and why I think it applies to visual data:

Information sources and transmitters

In Figure 1-1, the information source is the data source, or others’ reports formed from data sources, and you are the transmitter. You encounter many sources of information in the course of your job, whatever your role—everything from your email inbox to databases to your own experiences. You choose what information you pass on and to whom. This means you almost certainly need to filter or summarize that information in some way. You will definitely summarize or prepare the data if you are working directly with the data source. You’ll do this when working with data too—more about this in the next chapter.

Receiver and destination

In organizational communication, your receiver is likely to be the destination. You have probably learned which methods of communication are particularly effective for the people you work with. For example, if you send emails to your boss constantly but get no response, you will probably stop sending them and look for another method. Perhaps you will start by speaking to your boss directly. Direct conversation is a much easier way to ensure that your message is received, because you witness it happening—well, most of the time. I’m sure there’s been a time when you have spoken to someone directly, but they weren’t paying attention and therefore didn’t receive your message. In these cases, their body language will soon tell you whether you are being an effective communicator or not.

So why don’t we always communicate in person? Simply, we can’t, especially when working across different organizations or locations. The rise in remote working during the COVID-19 pandemic has shown the importance of in-person communication and how much harder it is to be heard remotely. After all, in a digital world you can’t just walk over to someone’s desk to ensure that the receiver gets the message you want them to. Video conferencing can help resolve some of those challenges. Still, too many video meetings can make it difficult to get someone’s time and attention.

Getting Through to Your Audience: Context and Noise

Understanding communication isn’t always easy, though. How many times have you been misunderstood? Hall describes how social context changes the way the audience decodes and interprets messages. The circumstances you are in when you receive a message makes a significant difference.

Imagine receiving a communication about average employee pay per grade. How you feel about your own pay would dramatically change the way you receive that information. If you receive less than the values mentioned, you would be unlikely to decode the message in the same manner as if you were paid considerably more than the values shown. The same would be true depending on how you grew up. If you come from a poorer background, you might be saddened by what you might consider excessive pay, especially to senior executives in large organizations.

Context is the background information and circumstances of a situation or event that help to provide meaning. Organizational culture, your location (in the main office, a branch, or remote), and seniority in the organization all play a part in setting the context for your work. To ensure that your receiver has the background information required to decode and understand your communication as you intend, you may need to provide additional context.

Let’s look at an example. I’ll use a mock retailer called Chin & Beard Suds Co. and its release of a new soap fragrance. You need to update the management team on sales. If you’ve sold 1,000 of the new product, you might be pleased and send a message that the product launch has gone well. Let’s look at some of the context and noise factors that might affect your message:

Experience

Team members might have been through many new product launches and have different expectations of what good progress means in terms of the number of sales.

Other messages

The receiver might hear other information that you don’t have. If the receiver hears that another product has to be dropped to be replaced by the new product, they might have different sales expectations if the older product sold in much higher volumes.

Market knowledge

If the receiver knows of an overall uplift in sales volumes for similar products, this might raise their expectations for the sales of the new product.

Some of the context you provide could include the following:

• Has the product met sales expectations so far?

• How has the product performed against its competition?

• What are customers saying about the new product?

Piecing all this information together is vital.

Another part of Shannon’s system is still applicable today and may be becoming an even bigger challenge: noise. Noise is not always literal sound (though it can be) but refers to any interference that affects the communication being received. Trying to talk with friends is a much harder task that requires more concentration in a restaurant with loud background music than in a quiet environment.

Ensuring that your message even reaches your audience can be a challenge when it is competing with many other messages for attention. The popular writer and statistician Nate Silver, in The Signal and the Noise (Penguin Press), defines noise as elements interfering with clear understanding of a communication. This can include having too many data points or communications (such as a constant stream of emails), unclear or overly technical language, difficulty meeting in person or online, and personality conflicts in meetings. Contrary opinions, audible or not, can cause confusion for the receiver: their knowledge and understanding of a subject will alter how they absorb the information you are providing. Knowing about your audience is key.

Finally, communication is successful when the receiver not only understands the information but retains it and incorporates it into their decision making. It has to be memorable. (After all, communication is usually about persuasion in some way.) Next, I’ll look at what we know about how the human brain retains information.

Don’t Forget About Memory

What does it mean to retain information, and how long do you need the receiver to remember your message? There are three types of memory. You’re likely to make use of them all:

Sensory

As the name suggests, sensory memory is triggered by your senses. When communicating visual information, the sense you are likely to trigger is visual. You are triggering a sensory memory if the information can be retained within a second. Can you quickly remember which months met the £208,000 profit target, which would mean the annual run rate would lead to a £2.5 million annual profit (Figure 1-2)?

Figure 1-2. C&BS Co. monthly profit levels

Just glancing at this chart, you will likely be able to see that the target is being met in only the later months of the year. When communicating with data, you’ll make particular use of a type of sensory memory called iconic memory, which stores visual information. This type of memory doesn’t last long, but it can help your audience remember key bits of information long enough to put a much more complex message into other types of memory.

Short-term memory

Short-term memory lasts from a few seconds to about a minute for most people. It can help the receiver build up more complex pieces of information in their minds, from multiple data points.

Research in the 1950s found that the average person’s short-term memory worked well for holding approximately seven items.¹ Newer research, however, suggests it might be only four items.²

You can enhance the length of your audience’s memories by using a technique called chunking, or breaking information into small chunks. Because you are aware of how many bits of information your audience can easily retain, you can optimize the amount of information you show them. This reduces the risk of overloading them.

Long-term memory

As you’d probably expect, long-term memory is thought to last up to a lifetime. When communicating with data, we actively call upon this type of memory less frequently. You can make use of your audience’s long-term memory by using themes that will remind them of long-held memories or information. There’s a reason family kitchens are used so frequently in television commercials: many people relate that location to memories they formed as children.

Next, I’ll show you how to use sensory memory to share key points through something called pre-attentive attributes. Without looking back at Figure 1-2, can you remember whether any months met the profit target? Hopefully, you can, and that is because of pre-attentive attributes that we will dive into deeper now.

Why Visualize Data?

Two words: pre-attentive attributes. This intimidating term simply refers to the ability to see patterns in images without having to think or consciously work to understand what you are seeing.

This ability evolved in humans to allow us to spot dangers, assess situations, and make instant decisions, without having to think about every little thing happening around us. For early humans, this was mostly finding food or avoiding being something else’s food, while today it might be more like seeing a car, a falling object, or a hazard in our path. We still use this part of our sensory system even when we’re not on the move.

Pre-attentive attributes can be used for more than just preventing danger. Data visualization relies on this pattern-spotting ability to communicate messages. By representing data in visual forms like bars, lines, or points, you can make use of pre-attentive attributes to grab your audience’s attention and make sure they receive your message.

What pre-attentive attributes can you use in data visualization? Figure 1-3 shows a sampling of the possibilities.

Figure 1-3. Visual representation of pre-attentive attributes

In this range of pre-attentive attributes, some are more effective than others. In Now You See It (Analytics Press), Stephen Few, an information technology innovator, highlights two in particular that humans are better at assessing precisely:

Length

Humans notice length at a glance, and we’re also good at estimating gaps between different lengths. We can use this to our advantage with data by showing the greatest values as the longest. Length is frequently visualized as a bar chart.

2D position

Often shown in the form of a scatterplot (a chart type we’ll explore in Chapter 4), 2D positioning places the greatest values at the top right of the chart. The 2D position is created by using two axes, one vertical and another horizontal. Comparing two metrics against one another is a common task in data analysis.

The other pre-attentive attributes aren’t assessed as precisely, but don’t disregard them. Precise comparison is not the only way to communicate data. For example, highlighting a key time period by using color or shape can capture your audience’s attention.

In an analysis about air pollution (Figure 1-4), I used size, color, and shape to grab the reader’s attention more than to communicate a precise message. The car visualization at the top is the first one you come across: it is designed to set the theme but also create intrigue. I used circles to demonstrate the volume of certain pollutants. The size of the circles increases with the ratio of the particulates in the air.

Figure 1-4. Visualization demonstrating nonprecise pre-attentive attributes

Look at the graphic for a moment and compare the size of the circles. Could you tell me the percentage difference between the largest orange circle and the second-largest orange circle? I know I can’t, and I made the visualization!

But that isn’t the point. I used orange to highlight the London borough of Camden and allow the reader to be drawn to the relevant metrics and compare them against the city’s other boroughs. It’s imprecise by design but still draws upon pre-attentive communication techniques. I was designing this view for a broad audience and therefore needed to use these techniques to share the insights I had found. Knowing what your audience will comprehend and how much work they are willing to put into decoding your message is a key factor in communicating well.

Pre-Attentive Attributes in Action

Let’s take a typical table of numbers and see how we can make its message clearer by using pre-attentive attributes. The table in Figure 1-5 shows the number of bikes sold in the first half of a year.

Figure 1-5. Table containing bike sales for stores in the United Kingdom

You are clearly an intelligent person (you have chosen to read this book, after all), so here’s a challenge. How many seconds do you think it will take you to answer the following questions?

• What is the largest value in this chart?

• How many stores beat their target of 450 bikes sold in a month?

• Which store’s sales fluctuate the most?

Did that take a few more seconds than you were expecting? It probably did, and it was probably slightly frustrating too.

The amount of effort a reader must use to interpret what they see is called cognitive load. You will come across this term a lot in this book; it is a key factor in measuring the effectiveness of your visualization choices. Making your audience think about what you are showing isn’t always a bad thing, but you need to make the cognitive load appropriate for what you are sharing. Tables often take significant cognitive effort to interpret.

So why do so many people in so many organizations still use tables of data to communicate the results of their analysis?

In “The “Right” Data”, I discuss the importance of capturing the questions your user might try to answer. Tables are a good fallback option for some audiences when you don’t know what your user might ask or be looking for in your data set. But we know the questions we want to ask of this table, so let’s look at how we could use pre-attentive attributes to make the answers easier to find. Let’s start with this question:

• What is the largest value?

• Answer: 989

We could use so many techniques to help answer this question. A simple change in color, as shown in Figure 1-6, is a particularly effective method and doesn’t require removing any other data points from the view. This approach has no subtlety but shows how effective highlighting can be to pick out a single value—in this case, the largest.

Figure 1-6. Highlighting the highest bike sales

Highlighting the highest value draws attention to it. When trying to find the highest value in a table of numbers, often you’ll be looking for the longest number (in terms of the number of digits), as that is likely to indicate the highest value. Here, the bike sales are all three digits long, so we need another method to draw the reader’s attention. To visually communicate more complex insights, we have many methods to choose from, depending on what you are looking to share.

But how could pre-attentive attributes be used to share other answers to questions your audience might have about the data in this table? Let’s consider the next question:

• How many times did stores beat their target of 450 bikes sold in a month?

• Answer: 17

This is a tough one! Without any visual clues, you are forced to read each number and assess whether it is greater or less than 450. You are not only assessing the value but also trying to count how many meet the target.

You could use a similar technique to the first question and just highlight the values that meet the condition set in a different color (Figure 1-7).

Figure 1-7. Highlighting values above the target (450)

But other methods might be more useful here. For example, using colored bars to highlight whether values fall above or below the target might make a simple count easier (Figure 1-8).

Again, the consumer of the chart will need to count the orange columns, but this is much easier than assessing whether a value is above the target first. To remove even the challenge of counting, you could create a chart just demonstrating that count (Figure 1-9), but you would lose the individual stores’ monthly sales values. As with any data communications, being specific about what question you are trying to answer can change how you visualize the data. More on that aspect in Chapters 3 and 4.

Once you solve the basic question, you might want to go further with your analysis. Other questions that could be asked of this data include the following:

• Which store’s sales fluctuate the most?

• Answer: York

Now we’re getting into some better analytical questions.

First, we must define fluctuating sales. I’ll use a simple definition: the greatest variation—specifically, the store with the largest difference between its best sales month and its worst.

Assessing this data using just values is really hard. As your questions become more complex, good use of data visualization will make finding the answers much easier.

Figure 1-8. Bike sales meeting the target

Figure 1-9. Count of stores beating their monthly sales target

My first instinct was to go back to our most effective pre-attentive attribute: length. Perhaps, I thought, a clear answer would appear if I drew a bar between the smallest sales value and the largest for each store. The result is Figure 1-10.

Figure 1-10. Sales variance shown as a Gantt chart

The bars in a Gantt chart don’t have to start at the zero point, but the chart still uses length to represent the value being shown.

Note

The Gantt chart is named after Henry Gantt, who designed this type of bar chart in the early 1910s. Gantt charts are often used as project-management tools.

However, this chart still isn’t the easiest to read: you have to pay close attention to where the bar starts. It’s much easier to remove the minimum and maximum values and just show the difference instead of the actual sales value (Figure 1-11). To make the analysis even easier for your audience, you could sort the stores from largest to smallest difference.

Figure 1-11. Sales variance shown as a bar chart

Now, seeing that the longest bar is York is much easier. Even though York has a similar sales variance to Leeds, having the bars start at the same place makes it much easier to spot the difference and interpret the chart.

In short, even when you’re using pre-attentive attributes well, you still need to take care that the chart conveys the information clearly, that you’re using the best pre-attentive attribute for the task, and that you’re keeping the question in the forefront of your mind. When you do, your message comes across clearly, without forcing the consumer to think too hard. If you don’t pay attention to these factors, you will create the opposite effect, and people will want to go back to tables.

Your understanding of pre-attentive attributes will help you make better choices about which charts will best communicate the message you want.

The challenges of communicating with data in your organization are likely to resemble the challenges of any other form of communication. You will need to find ways for your communication to be received, decoded, and remembered. This book will help.

Unique Considerations

What makes data communications different from other kinds of communication? First, as you might guess, they are based on a data source or data analysis. (You’ll learn more on finding the right sources in Chapter 2.)

Second, data communication in the workplace is usually about meeting a stakeholder’s requirement or answering a question. You’ll need to know what those requirements and questions are and then analyze your data to find the answers.

Third, data communication, as covered in this book, is all about visual analysis. Chapters 3 and 4 will show you lots of options to analyze your data visually. The type of chart you choose is ultimately the signal you are sending, so you’ll need to learn how to make the right choices for your audience.

Fourth, data communication is about trust. Your points carry more sway when you can show how they are supported by evidence. If data-informed decisions have failed in the organization before, you will need to work hard to build trust. Building trust with the receiver of your communications will reduce the noise of other opinions or messages that don’t have the same level of supporting evidence. You need to be confident that your message will be heard; when it is, you can influence more decisions and get more done.

You will need to build trust in your data analytics skills. It’s easy to manipulate data to support an agenda—filter heavily enough, ignore outlying data points, or use other tricks and you can eventually get the data to say what you want it to. Whenever data is used to support a political point or marketing campaign, you should look to the source to see how the data may have been manipulated. For your own work, your audience needs to know that you are showing a fair representation of the data points from the data source you are using. This level of trust will build as you continually provide fair, well-sourced, and useful data-based communications. As you spend time exploring data sets to see what stories are held within, remember to share what you were expecting to find as well as what you actually found. Telling a balanced story only enhances the weight of your opinion.

Another factor is trust in the data sources themselves. The rise of modern self-service data tools that focus on visualization has made it much easier for nonspecialists to access and work with data and to ask and answer important questions about its provenance and reliability (more on this in Chapter 2). If that’s why you picked up this book, you’re in the right place. Chapters 8 and 9 will get deeper into the specifics of how to help your audience understand and trust your data.

Chapter 2 will go over the fundamentals of working with data—what it is, what you need to do with it, and what’s so important about it. Chapters 3 and 4 focus on the practical aspects of visualizing data, such as formats and chart types, and contrast traditional approaches with more innovative ones. Chapters 5 and 6 build on this by teaching you visual techniques for clarifying your communication. The communications philosopher Marshall McLuhan famously said that “the medium is the message,” and Chapter 7 is a deep dive into how the medium and format you choose influence your audience and the way they receive your message. Finally, Chapters 8 and 9 get into the nuts and bolts of putting all this to work in a real organization full of real people with different needs and interests. Chapter 8 looks at the challenges of communicating in the workplace generally, while Chapter 9 zooms in on specific types of departments and teams for a practical discussion of their communication needs.

Summary

Communication is key to getting anything done at work. You will need to be clear so your audience can receive, decode, and remember your message. Making use of pre-attentive attributes will help you do this more effectively.

Data can help you communicate your message more clearly and explore the evidence for and against the points you are making, so you can confirm your ideas with evidence—or adjust them to fit the evidence—before sharing them with others.

¹ G. A. Miller, “The Magical Number Seven Plus or Minus Two: Some Limits on Our Capacity for Processing Information,” Psychological Review 63, no. 2 (March 1956).

² N. Cowan, “The Magical Number 4 in Short-Term Memory: A Reconsideration of Mental Storage Capacity,” Behavioral and Brain Sciences 24, no. 1 (February 2001).